An Extended Project Qualification (EPQ) research project that applies machine learning to predict agricultural crop yields using historical satellite telemetry and weather data. The project focused on avocado crop yields in California as a case study.
- 70.4%
- Model Accuracy
- on test data
- 2+
- Data Sources
- integrated
- 18
- Months
- of research
- StackPython, TensorFlow, Scikit-learn, Pandas, NumPy, Seaborn, Matplotlib
- PlatformCross-platform (Windows/macOS/Linux)
- GitHubEPQ Crop Yield AI Repository
Data Sources
- AgriculturalFAOSTAT - Food and Agriculture Organization
- ClimateWorld Bank Climate Knowledge Portal
- ResearchSatellite Yield Estimation Research
Methodology
The project involved collecting and preprocessing historical satellite imagery data, weather patterns, and agricultural yield records. Multiple machine learning models were trained and compared, including neural networks and ensemble methods, to find the optimal approach for yield prediction.
Key Learnings
- • Feature engineering from satellite imagery (NDVI, land surface temperature)
- • Time series analysis for seasonal agricultural patterns
- • Model selection and hyperparameter optimization
- • Data visualization and scientific communication
Reference Material
This tutorial video was instrumental in developing the machine learning pipeline for this project.
